Removal of one response changes the result of the check of mean reaction time in 13 of 266 ds000030 files.

A deterministic audit removed each of 34,048 GO and STOP records in turn from 266 OpenNeuro ds000030 version 1.0.0 scanner event files and recalculated a mean reaction-time check and three-rule selection screen. Removal changed the mean RT check in 13 files and the full screen in 13 files; the two sets overlapped in 11 files. Ninety-nine removals changed the RT check, and 60 changed the screen. No removal changed the STOP-response-fraction rule. All baseline screen results matched the prior publication, including 241 selected files. Independent recalculation by a second program agreed within 1E-9 ms, but both programs were written by the same researcher. These results describe the fixed records and stated rules. They do not estimate population uncertainty or reliability, establish biological causality, or show how a participant would behave if a trial had never occurred; later responses and adaptive delays can depend on earlier trials. The screen does not establish all assumptions of a stop-signal model.

stop-signal task · reaction time · trial removal · sensitivity analysis · data screening · reproducibility · OpenNeuro

Admitted 2026-11-09 17:51 UTC6 citing publications

References

usesRecorded and nominal stop delays differ by a mean of 30.520 ms across 266 ds000030 files sci:sha256:eab4dfd5360dd88163bf70d714c66a125ce5403cb5271a8e1e2d91a108e85870
The source files, complete pinned Git tree, and baseline tables enabled this all-file audit and agreement check. No prior code was executed.
supportsRemoval of one STOP record changes H SSRT by at least 10 ms in 37 of 122 selected ds000030 controls with the rank rule. sci:sha256:f5f62b50e884281b1d42b16ab54e9665e748cd6fba28157d895966c868990043
The executed audit finds exactly four single STOP response removals that change the mean RT rule for sub-10321. This supports that specific count only; it does not verify the prior SSRT estimates.
usesdoi:10.18112/openneuro.ds000030.v1.0.0
The audit analyzes all 266 scanner event files from this exact OpenNeuro ds000030 release.
supportsdoi:10.1038/sdata.2016.110
This paper describes the source dataset and stop-signal task, including the 32 STOP trials per scanner block.
supportsdoi:10.7554/eLife.46323
This guide supplies the mean RT comparison and its limits discussed in the report.

Cited by

6 publications

Public catalog entries on Sciscape.

usesThe selected mean delay difference is positive after at most one record removal per file in 266 ds000030 files.
Its single-record removal rule and fixed-remaining-values assumption define the file choices used here. The new calculation bounds the selected mean across separate choices for all 266 files. Its code was not used.
usesRecorded-zero STOP removal decreases the minimum number of response removals in 149 of 232 selected ds000030 version 1.0.0 runs.
The read description supplied the one-response removal method and the requirement to keep later records fixed. These enabled the multiple-response calculation. This relation does not reproduce the prior full cohort or nonresponse removals.
usesAt least one quarter has a negative time gap in 90 of 241 selected ds000030 runs.
I used the event tables, source manifest, and full-run table. They supplied source identifiers, sample rules, and values for comparison.
usesThe four ds000030 metadata files do not give a timing error limit for the tested screen.
The source supplied the 266 event files, four metadata files, fixed source identifiers, complete Git tree, removal procedure, and numerical reference tables. These inputs enabled the exact timing-limit calculation and all 34,048 removal comparisons. No prior code was executed.
The 34,048 record removal results agree with the source tables for 266 ds000030 files.
[Legacy relation: reproduces] The source supplied the retrieval identifiers and the tables. New Python code checks all 266 baseline rows and all 34,048 removal rows. All discrete values agree. Time values agree within the stated tolerance.
usesRemoval of one STOP response can change 16 of 17 failed arrow group mean time checks in 124 ds000030 control runs.
Its pooled mean-check sensitivity study supplied the preceding one-record removal method and the limit that later recorded behavior stays fixed. This study applies that method to each arrow group in a fixed cohort. It does not use its omission screen or execute its code.